What do I do when I'm stuck?
Guessing alone is how a six-week build turns into a six-month one. Here, being stuck has a defined path: people who have shipped this kind of system and see your work every week, a coach that already knows what you are worried about, and a set of commands built for the exact moment you stall.
Who You Escalate To
The instructor and mentors behind Agentic Buildcamp are practitioners who have shipped things other people depend on. They push on your build the way a real skeptic would, while there is still time to fix it.
Stand-ups, daily in the first and final weeks
Short live sessions with your mentor and cohort where you say what is blocking you and get an answer on the spot. Hearing what everyone else is stuck on is half the value: most blockers are shared.
Slack between sessions
Async mentor support runs through the whole program. You post the actual error, the actual design question, and a practitioner who has shipped this kind of system answers.
Cycle demos, reviewed individually
Every recorded demo you submit is reviewed by a mentor, and a few are discussed by the whole group. The feedback is about your build specifically, and it arrives while you can still act on it.
1-on-1 mentorship, if you add it
Weekly calls on your project alone, booked directly against your mentor's availability. Personal training on top of the group session, for builders who want a second pair of eyes on every decision.
AMIR FEIZPOUR
Dr. Amir Feizpour is the Founder, CEO, and Chief Scientist of Aggregate Intellect, where he is building a generative business brain for service- and science-based companies. He has grown a global community of over 6,000 AI practitioners and researchers focused on AI research, engineering, product development, and responsible AI.
Previously, Amir was NLP Product Lead at the Royal Bank of Canada and held a research position at the University of Oxford, conducting quantum computing experiments that led to high-profile publications and patents. He earned his PhD in Physics from the University of Toronto.
Amir also contributes to the AI ecosystem as an advisor at MaRS Discovery District, entrepreneur-in-residence at North Forge and Lab2Market, and fractional Chief AI Officer for several startups. He leads Aggregate Intellect's R&D collaborations with McGill University, the University of Toronto, and the Alberta Machine Intelligence Institute, exploring the future of knowledge work.
Mentors answer your implementation questions asynchronously, or synchronously on 1 on 1 calls, and you can also watch them build and learn their thought process, techniques, and favorite tools.
Azadeh Mostaghel
Azadeh Mostaghel is a systems-focused engineer and startup founder with deep expertise in building scalable, intelligent...
More
Invited Participants
We often invite a small number of our alumni that show a great deal of enthusiasm to build and excel at knowledge sharin...
More
Your Personalized Journey
Getting unstuck is personal: the right next step depends on what you were just doing, what is worrying you, what other people have told you, and how much runway is left. Sherpa-B keeps all of that in view so its advice is about your project today.
Your worries are on file, and the coach reads them back
Every
/coachsession ends by updating a builder priorities checklist in your own words: what you were unsure about, dated. The next session opens by asking whether that is still on your mind. A still-open concern visibly moves related Tasks up or down the recommendation instead of getting a passing mention.What other people said carries its own weight
A separate feedback checklist holds what personas, users, mentors, and demo audiences actually said, each entry tagged with its source. The coach weighs "a CTO sparring round flagged this" differently from "two people in your last demo asked about this," and calls it out when the same gap shows up from more than one source.
Timing changes the advice
The coach knows the cohort calendar and the time of day. Starting at 9pm with a demo in two days gets a different recommendation than a fresh Monday with two weeks of runway.
Stuck on a concept? A third terminal, one topic at a time
When you cannot place a concept, judge an output, or understand a tool, the coach hands you a specific research prompt for a dedicated learning terminal, separate from both the coaching session and the work. Come back when you have it. It also asks you to explain what happened in your own words before confirming or correcting, so the fix sticks.
The Stuck Toolkit
These are the commands and workouts you run in your coding agent the moment you stall, and the ones that turn every demo and conversation into signal you can act on.
Coach
/coachThe long-lived coordinator you keep open in one terminal. It gathers your Tasks, issues, calendar, and both checklists, recommends what is next with a reason, hands you the exact command for the work terminal, then verifies the work actually happened (a commit, a changed file) before closing the Task out.
Sparring Session
/sparring-sessionRehearse the objections you have not heard yet. One persona, one round, sized to the time you have: nine to choose from, including a customer, a CTO, an investor, and four kinds of hiring manager. Tap out any time. The log and open items feed the shared feedback checklist.
Capture Feedback
/capture-feedbackAfter a demo, a user-testing session, a mentor chat, or a round of social reactions, you paste the raw notes and talk through what they actually mean before anything gets filed: what surprised you, whether it contradicts your spec, whether it is a pattern or one loud opinion. Then it files issues, updates the checklist, and flags any assumption that needs revisiting.
Demo Prep and Submission
/sprint-demo-prep/submit-demoBefore a cycle demo, one command reads your recent commits, reflections, and open feedback items, then drafts a timed four to five minute script that addresses known gaps head-on instead of getting caught by them live. The other checks your share link actually opens for a stranger and submits it for mentor review.